An Experimental/Computational Approach

to Identify Moduli and Residual Stress

in MEMS Radio-Frequency Switches
by H.D. Espinosa, Y. Zhu, M. Fischer, and J. Hutchinson
ABSTRACTIn this paper, we identify the Youngs modulus
and residual stress state of a free-standing thin aluminum
membrane, used in MEMS radio-frequency (rf) switches. We
have developed a new methodology that combines a mem-
brane deection experiment (MDE) and three-dimensional
numerical simulations. Wafer-level MDEtests were conducted
with a commercially available nanoindenter. The accuracy
and usefulness of the MDE is conrmed by the repeatabil-
ity and uniformity of measured loaddeection curves on a
number of switches with both wedge and Berkovich tips. It was
found that the loaddeection behavior is a function of mem-
brane elastic properties, initial residual stress state and cor-
responding membrane shape. Furthermore, it was assessed
that initial membrane shape has a strong effect on load
deection curves; hence, its accurate characterization is crit-
ical. Through an iterative process and comparison between
MDE data and numerical simulations, the Youngs modulus
and residual stress state, consistent with measured mem-
brane shape, were identied. One important nding from this
investigation is that variations in membrane elastic properties
and residual stress state affect the loaddeection curve in dif-
ferent regimes. Changes in residual stress state signicantly
affect the loaddeection slope at small values of deection.
By contrast, variations in Youngs modulus result in changes in
loaddeection slope at large deections. These features are
helpful to decouple both effects in the identication process.
KEY WORDSMEMS, rf switch, moduli, residual stress,
nanoindentation
Introduction
Microelectromechanical systems (MEMS) are one of the
most signicant technological advances of this decade. This
technology employs systems with dimensions of only a few
hundred micrometers. Applications ranging from drug de-
livery systems to telecommunications are currently under
research and development. Their reduced size and mass give
them unique technological advantages. For instance, radio-
frequency (rf) switches used in wireless applications exhibit
microsecond time responses, low insertion losses and low
H.D. Espinosa (espinosa@northwestern.edu) (SEMMember) is a Professor,
Y. Zhu is a Graduate Student, and M. Fischer is a Graduate Student,
Northwestern University, Evanston, IL 60208-3111. J. Hutchinson is a
Professor, Harvard University, Cambridge, MA 02138.
Original manuscript submitted: December 19, 2002.
Final manuscript received: December 19, 2002.
power consumption. However, size reduction also results in
mechanical characterization challenges.
The impact of the technology is such that during the rst
few months of 2001 three feature articles appeared in the
IEEE Spectrum
1,2
and Mechanical Engineering magazines.
3
In all of these articles, the problem of MEMS reliability, due
to lack of fundamental understanding of failure mechanisms,
is highlighted as a major limitation for the commercialization
of MEMS devices. Key problems to be solved are packaging
and mechanical modeling of MEMS materials at the microm-
eter scale. For instance, in the case of rf switches, the effect
of the environment can result in stiction of the membranes
due to humidity or other sources. From a reliability point
of view, it is necessary to consider the plasticity limit and
its dependence with temperature of the materials involved.
Temperatures as low as 50

C can be reached in satellite and

airplane wireless applications while temperatures of a few
hundred degrees can be present during packaging of the de-
vice. Another possible failure mechanism is fatigue due to a
large accumulation of actuation cycles. Most of these devices
are typically actuated at very large number of cycles pushing
the design envelope and our current knowledge of material
behavior beyond known parameters.
In this paper, we address the identication of the Youngs
modulus and residual stress state of free-standing membranes
used in the design of MEMS rf switches. We begin with a de-
scription of the MEMS rf switch investigated in this study.
Then, limitations of currently available experimental tech-
niques are discussed and a new test is presented. A subse-
quent section refers to three-dimensional numerical simula-
tions carried out to identify the device residual stress state
and the material Youngs modulus. The initial shape of the
free-standing membranes is taken into account. Finally, we
give conclusions arising from the combined experimental
computational methodology and its applicability to the study
of MEMS reliability.
The MEMS rf Switch
Goldsmith et al.
47
have reported on MEMS rf switches
employing free-standing metal membranes, with capacitive
coupling. The side-view architecture of one such MEMS
switch is shown in Fig. 1(a). The switches are manufactured
on a GaAs substrate, over which a bow-tie metal mem-
brane is deposited by evaporation. The material used in the
membrane is an aluminum alloy. This membrane is the only
moving part of the device. Its shape, size and mechanical
2003 Society for Experimental Mechanics Experimental Mechanics 309
G
H
SUBSTRATE
MEMBRANE
RF IN
RF OUT
BOTTOM
ELECTRODE
BUFFER LAYER
Fig. 1(a) Cross-section of MEMS rf switch. (b) Optical
micrograph of the bow-tie membrane mounted on posts.
The membrane is made of aluminum alloy and contains
a pattern of holes for membrane release during plasma
etching. All dimensions are in micrometers.
properties determine the behavior of the MEMS switch. Two
of its edges are attached to thin posts that maintain it sus-
pended over an insulated electrode. Microfabrication details
and employed materials have been reported by Goldsmith
et al.
4,6
Switching is achieved by applying a pulling-in volt-
age between the membrane and the bottom electrode.
Figure 1(b) shows the top view of a switch together with
its dimensions. The membrane thickness is 300 nm with a
variability of 10 nm from membrane to membrane. The
gray circular dots over the membrane are small holes, 2 m
in diameter, which are necessary for plasma etching of the
polymersacricial layer. These holes also play a role in the
dynamic behavior of the switch, by providing viscous damp-
ing, if actuated in an inert gas or air.
Methodology
A review of the literature was carried out to identify an
experiment suitable for the identication of the membrane
Youngs modulus and residual stress state. Among the most
established experimental techniques, we can cite the bulge
test,
810
the microtensile test developed by Sharpe et al.
1114
and nanoindentation,
1519
especially the deection of circu-
lar membranes.
20
The bulge, microtensile tests and circular
membrane deection techniques require the microfabrication
of special specimens. In view of the fact that the metal mem-
brane grain structure and residual stress state are highly de-
pendent on the substrate material and fabrication steps, the
usage of specially designed samples was not deemed appro-
priate. Likewise, we found indentation tests performed on
areas of the membrane on top of the posts very unreliable.
Strong depth dependence was measured and the modulus
and hardness never reached a constant value, see Fischer
21
and Espinosa et al.
22
The unsuitability of these experimental
methodologies for the device under investigation prompted
us to develop a new combined experimentalcomputational
methodology.
A wafer-level experimental technique was employed to-
gether with three-dimensional numerical simulations of the
experiments.
23
The overall experimental/computational ap-
proach consisted of:
membrane characterization, to measure dimension and
topography using optical full eld prolometry;
membrane deection experiments (MDEs), to interro-
gate the structural response of the device;
three-dimensional nite element analysis (FEA), to
correlate the experimentally measured loaddeection
curves.
Makinguse of a nanoindenter (MTSCo.), MDEs were per-
formed to investigate the membrane structural response. The
technique should not be confused with nanoindentation or
with microbending of cantilevers or xedxed beams.
17,24
Since the sample is attached to the posts along two of its
edges, and a gap of about 4 m exists under the membrane,
a loaddeection measurement was used to interrogate the
membrane elastic response. The experiment consisted of ap-
plying either a point or a line load in the middle of the span
of the xedxed membrane. A specially designed 125 m
long diamond wedge tip and a Berkovitch tip were used to
apply the loads. The idea of using two tip geometries was
to examine the sensitivity of the approach to load geometry
as well as to obtain redundant experimental information. By
deecting the membrane down, information on both elastic
behavior and residual stress state was obtained. Obviously
the structural response was coupled, so analysis was needed
to decouple the effects.
The experiment was quite challenging because the stiff-
ness of the specimen, about 10 Nm
1
, is only 10%that of the
nanoindenter column stiffness in air, which is approximately
100 Nm
1
. Furthermore, thermal drift is part of the measured
load signature. However, the nanoindenter is capable of ac-
quiring load and displacement with high resolution and to
perform several pre-programmed steps on the same or differ-
ent membranes. Hence, we were able to develop a procedure
for removing spring and thermal drift effects in the load sig-
nature, which proved to be quite robust and accurate.
21,25
Loads with a resolution of a fraction of a N and displace-
ments with a resolution of less than a nanometer were so
measured.
The protocol normally used in nanoindentation tests was
employed in these MDEs. Two positioning indentations, over
the posts, were rst made to measure device tilt, membrane
height in the middle of the span, and to accurately determine
the center load location. This positioning was carried out with
310 Vol. 43, No. 3, September 2003 2003 Society for Experimental Mechanics
an accuracy of less than 1 m, which is less than 0.32% of
the membrane span. The device tilt was adjusted until the
indentation marks left by the wedge tip, over the posts, had
uniform depth. A thermal drift check, below a desired value,
was performed before the start of the loading phase.
The set of parameters employed in the experiments was:
surface search distance = 6000 m, radius to initial impact =
50 m, angle to initial impact =80

, maximumdrift rate prior

to tests = 1 nm s
1
, and tip displacement rate = 10 nm s
1
.
The surface search distance is the distance above the sample
at which the approach stage begins. For these membranes,
a long approach (6 m) was used due to surface curvature.
A large number of data points, during the approach stage,
were recorded in order to be able to characterize the change
in stiffness when the tip contacted the membrane. The tip
approach rate was 60 nm s
1
until it reached the bottom
electrode, where the tip displacement rate was changed to
10 nm s
1
to avoid large load increases that could damage
the device.
The FEA analyses were carried out in two steps. The rst
step assumed values of Youngs modulus, uniform uniax-
ial stress and a rst approximation to the membrane initial
shape. The solution of this step resulted in an equilibrium
non-uniform stress state and a compatible membrane geom-
etry, which was compared to the measured one. This rst step
was performed iteratively until the computed and measured
membrane geometries agreed to within a set error. The er-
ror norm used in the calculations is reported in the following
section. The second step simulated the motion of the nanoin-
denter tip and computed the interaction force between tip and
membrane. A loaddeection response was so computed.
Results and Discussion
The rf MEMS Membrane Metrology
Prior to the MDE tests, a careful characterization of the
membrane dimensions and out-of-plane initial shape was car-
ried out. The in-plane dimensions of the investigated rf switch
are given in Fig. 1(b). The membrane presents a bow-tie shape
necessary for optimal switch performance. The total mem-
brane spanis 316mandits widthvaries from175to115m.
The hole pattern and diameter of 2 mwere also measured to
include them into the solid model employed in the numerical
simulations.
The membrane surface was scanned using the MicroXAM
3D surface proler system. The measured membrane topog-
raphy exhibited waviness along the span in both directions.
The measured membrane shape can be mathematically de-
scribed by the following function
Z(x, y) = Acos

5x
L

cos

0.8x
L

cos

y
W

, (1)
where A is the amplitude, L is the length of the membrane
and W is the width. The average amplitude was found to
be 0.2 m. Measurements on several membranes show sim-
ilar topography with minor variations in the magnitude of
the curvatures. As will be shown later, this initial membrane
shape plays an important role in the proper identication of
membrane mechanical properties.
A focused ion beam (FIB) was employed to characterize
the lm material grain size and membrane thickness. An im-
age taken froma sectioned membrane is shown in Fig. 2. The
Fig. 2Image showing the grains of the aluminum alloy
membrane after removing a top thin oxide layer. The bottom
oxide layer is observed as a thin lighter border along the edge.
The sectioning was performed using a FIB. The average
grain size is about 250 nm and the membrane thickness is
300 nm with only one grain through the thickness.
micrograph shows grain morphology and size immediately
after sectioning and etching of a native oxide layer. A 20 nm
thick native oxide layer can be observed in Fig. 2. A single
grain through the thickness with a columnar morphology is
also clearly observed. The average grain size is about 250 nm.
Note that the sample is tilted 45

to better image the grains.

Membrane Deection Experiments
Figure 3(a) shows experimental results, obtained by test-
ing one switch through repeated loading. The four tests were
conducted employing a wedge tip. Contact with the mem-
brane is taken as the zero displacement. As previously ex-
plained, the membrane response to stretching was obtained,
as it was pulled down by a displacement equal to the gap.
When the bottomelectrode was reached, we observed a steep
increase in load, corresponding to a large stiffness increase.
The membrane was not damaged as a result of the test, and
no mark was found after it. The repeatability of the test is
clearly demonstrated in view of the small variations in the
recorded loaddeection curves.
The same deection experiment was conducted on four
different switches to examine wafer uniformity. The mea-
sured loaddeection curves are shown in Fig. 3(b). A very
small spread in the experimental curves is observed, indica-
tive that the microfabrication process is quite uniform over
the wafer. Small variations in maximum deection prior to
contact with the bottom electrode are also observed, which
is consistent with the variation in membrane initial shape.
Similar experimental results were obtained when a con-
centrated force, using a Berkovich tip, was applied at the
center of the membrane. The loaddisplacement signatures,
Figs. 4(a) and (b), again conrm the test repeatability and
switch uniformity across the wafer. In all cases very small
unloading events are recorded along the deection process.
It was determined, by examining the nanoindenter column
motion in air, that such perturbations may be the result of the
nanoindenter column behavior rather than true device behav-
ior. However, another possibility is slipping of the tip as a
result of membrane local deformation associated to its initial
2003 Society for Experimental Mechanics Experimental Mechanics 311
(a)
(b)
Fig. 3(a) MDE curves obtained from a single membrane
repeatedly tested using a wedge tip. Repeatability is ob-
served. (b) MDE curves on four different membranes using a
wedge tip. Small variability among rf switches is observed.
waviness. It should be noted that the instrument is used at the
limit of its resolution.
Numerical Simulations
The membrane deection experiment was simulated using
the nite element software ABAQUS Standard, version 5.8.
The simulation consisted of a quasi-static three-dimensional
contact problem with the tip modeled as a rigid body and
the xedxed membrane as a shell attached along its edges,
i.e., Z(L/2, y) = 0. The edges of the membrane along the
posts were xed to simulate the bonding between membrane
and posts. The membrane was discretized using eight-noded
shell elements with variable dimensions in the xy plane
and a constant thickness of 300 nm in the z-direction. The
element type used to model the membrane was the S8R5.
The diamond wedge tip was modeled as a rigid body with a
tip radius of 40 nm and 80

incline angles. The element type

(a)
(b)
Fig. 4(a) MDE curves obtained from a single membrane
repeatedly loaded using a Berkovich tip. Repeatability is
observed. (b) MDE curves on four different membranes using
a Berkovich tip. Small variability among the rf switches is
observed.
R3D4 was employed in discretizing the tip. The rigid element
nodes were master nodes in the contact pair while the shell
elements dened the slave surface. The mesh is shown in
Fig. 5 for both loading types. The gure is a zoom of the tip
and membrane. Avery ne mesh is used to dene the contact
area under the tip.
Nonlinear kinematics was included in the simulations to
examine the changes in membrane stiffness with deection.
Material plasticity was not accounted for since, at maxi-
mum deection, the stress state everywhere in the membrane
was below the material yielding stress. The numerical anal-
ysis was performed in two steps. The rst step consisted of
solving the system of governing equations for an initial uni-
form uniaxial stress and a rst approximation of the initial
membrane shape. Equilibrium resulted in a new membrane
312 Vol. 43, No. 3, September 2003 2003 Society for Experimental Mechanics
Fig. 5Plot of line (wedge tip) and point load (Berkovitch tip) nite element meshes in the contact region. A rened mesh was
used in the contact area to capture local bending and tip geometry.
shape and a non-uniform biaxial stress distribution. The sec-
ond step consisted of applying a prescribed displacement to
the tip. The prescribed displacement rate was the same as
that used in the experiment, although this was not necessary
considering that the simulation was quasi-static. The contact
load was an outcome of the calculation dependent on the
membrane behavior.
In order to analyze the sensitivity of the membrane struc-
tural response to variations in Youngs modulus, the model
was examined with four values of E at a xed initial uniaxial
stress of 10 MPa. A at membrane was used in these simu-
lations. The bulk modulus of aluminum, 73.2 GPa (10.6
10
6
psi), and three smaller values, 50, 45, and 40 GPa, were
examined. In all calculations, a Poissons ratio of 0.34 was
used. The simulated loaddisplacement behaviors are shown
in Fig. 6(a). Considering that a variation of 33.2 GPa in E is
examined, the differences in loaddeection curves are mod-
est, and almost negligible at small values of deection. By
contrast, the slope of the loaddeection curve varies sig-
nicantly, for the various values of E, when the deection
reaches 3 m.
To examine the effect of initial stress,
0
, on the load
deection curve, four simulations were performed with ini-
tial stress of 0, 5, 10, and 15 MPa. A xed Youngs modulus
of E = 45 GPa and a at membrane geometry were em-
ployed. The simulated loaddisplacement curves are shown
in Fig. 6(b). It is interesting to note that the initial slope of
the loaddeection curve is highly sensitive to the state of
initial stress. In fact, higher values of initial stress translate
in a tension stiffness that becomes evident even at small de-
ections. As mentioned in the previous paragraph, variations
in Youngs modulus have little effect in the stiffness of the
membrane, when the deections are small. These sensitivity
features are very important to properly identify membrane
moduli and initial residual stress, as it is the objective of this
investigation.
The previously discussed analyses were all performed us-
ing a at membrane. This is enough for the purpose of un-
derstanding the effects of E and
0
through the identication
process. However, due to thermal and lmdeposition effects,
the membranes in each switch have an initial wavy geometry.
For the purpose of investigating this effect, several membrane
topographies were analyzed, at and wavy with two different
G
H
Fig. 6(a) Effect of Youngs modulus on load-deection
response. (b) Effect of residual stress on load-deection re-
sponse. The membrane shape is at in all these simulations.
2003 Society for Experimental Mechanics Experimental Mechanics 313
Fig. 7Effect of initial membrane shape on load-deection
behavior. The shape function for the wavy shape is shown in
eq (1). For wavy shape 1, the amplitude is 0.2 m; for wavy
shape 2, the amplitude is 0.4 m.
amplitudes (0.2 and 0.4 m). The loaddeection curves for
these cases are shown in Fig. 7 as calculated for an initial
stress of 27 MPa. Signicant changes in loaddeection be-
havior are observed with the at membrane geometry being
the stiffer, as would be expected. The effect of the ampli-
tude of the waviness is illustrated when wavy shapes 1 and 2
are compared. These simulations demonstrate that the model
must accurately capture the initial shape of the membrane.
The above sensitivity study shows that the identication
process involves three variables: initial membrane shape, ini-
tial stress, and Youngs modulus (S,
0
, E). While the effects
of
0
and E can be decoupled, by examining the stiffness
of the membrane at zero and maximum deection, the initial
membrane shape affects the membrane stiffness in both de-
formation regimes and, therefore, the identication of both
parameters. As a result we have performed the identication
of (S,
0
, E) through an iterative process.
We next report on the convergent solution for both steps
of the analysis. The rst step of the FEA analysis was carried
out by assuming a Youngs modulus (E), an initial assumed
shape (S) and a uniaxial uniform stress (
0
). Clearly this
initial state did not satisfy equilibrium. By solving the gov-
erning equations for the discrete system, a solution with a
non-uniform biaxial stress distribution and a new membrane
shape was obtained. As a criterion of solution convergence,
we employed an error norm given by
||e|| = [
n

i=1
(Z
cal
Z
meas
)
2
]
1/2
/Z
max
. (2)
Here, Z
cal
is the numerically calculated Z coordinate of
the membrane, Z
meas
is the measured one given by eq (1),
Z
max
is the maximum measured Z coordinate and n is the
number of mesh nodes.
Cauchy stress contours of S
xx
and S
yy
resulting from this
analysis are shown in Figs. 8(a) and (b), for the case of an as-
sumed uniforminitial uniaxial stress of
0
= 27 MPa. For this
calculation, an error ||e|| < 3% was used. A large variation
in calculated residual stress is observed as a function of posi-
G
H
S22 (MPa)
>
?
SJJ (MPa)
>
?
Fig. 8(a) Contour plot of Cauchy stress S11 in the local
11 direction for E = 75 GPa and uniform initial stress
0
=
27 MPa. (b) Contour plot of Cauchy stress S22 in the local
22 direction. Due to the symmetry of membrane geometry
and loading, only a quarter of the membrane is shown. Note
that the local 11 direction is the projection of the global
X-axis onto the membrane surface, and the local 22 direction
is at right angles to the local 11 direction. In most of the
membrane except for the edge area, S11 is between 0 and
6 MPa, and S22 is between 3 and 3 MPa.
tion. In the relatively at part of the membrane, the average
residual stress in the X-direction is between 0 and 6 MPa. By
contrast, close to the xed ends, compressive stresses as high
as 24 MPa are observed. The average residual stress in the Y-
direction is between 3 and 3 MPa in the relatively at part.
These ndings are in agreement with the non-uniform stress
distribution reported by Chen et al.
26
for the same MEMS
device.
The membrane shape and state of residual stress computed
inthe rst stepwas usedas initial conditionfor the secondstep
of the FEA simulation. This second step consisted of apply-
ing a prescribed displacement to the nanoindenter tip while
computing the contact force. The loaddeection curves re-
sulting from this second step, for both line and concentrated
loads, are shown in Fig. 9(a). The good correlation between
simulations and experiments proves that the solution set, E =
75 GPa, computed initial shape and associated non-uniform
biaxial stress state, provides an accurate mechanical descrip-
tion of the MEMS device. An important feature to note is that
the same set of parameters precisely captures loaddeection
signatures obtained with two different loading geometries.
This provides condence in the accuracy of the experimental
measurements and identied parameters.
Fig. 9(b) shows the comparison between two simulations
and the four experiments reported in Fig. 3. The simulations
are for line load, with the upper curve for E = 80 GPa and the
lower curve for E = 70 GPa. It is seen that the experimental
314 Vol. 43, No. 3, September 2003 2003 Society for Experimental Mechanics
(a)
(b)
Fig. 9(a) Comparison of numerical simulations with
experiments for both line load and point load for E = 75
GPa and uniform initial stress
0
= 27 MPa. (b) Comparison
of simulations with four experimental results. The upper
solid line was obtained with E = 80 GPa and the lower solid
line was obtained with E = 70 GPa. Both calculations were
performed based on the same membrane shape and residual
stress state.
curves lie between two simulations, which implies that E =
75 5 GPa.
Conclusions
We have developed a procedure for the identication of
the Youngs modulus and residual stress state in thin alu-
minum membranes, employed in the design of MEMS rf
switches. The strong dependence of these quantities on mem-
brane shape, attachment to the substrate and microfabrication
steps, severely restricted the selection of a suitable testing
methodology. Microfabrication of specimens for bulge or mi-
crotension tests was ruled out for these reasons.
It was determined that the loaddeection response of
these membranes is very sensitive to residual stress at small
deections. By contrast, the sensitivity to variations in the
Youngs modulus becomes important at large deections.
These ndings have the implicationthat the effects of Youngs
modulus and residual stress state can be decoupled. Further-
more, it was shown that proper identication of Youngs mod-
ulus and residual stress state requires that the initial shape of
the membrane be taken into account. In our methodology,
this was accomplished by performing membrane metrology
and a two-step FEA analysis.
The MDEtest provedtobe veryreliable evenwhena mem-
brane stiffness of one order of magnitude smaller than the
nanoindenter column stiffness was measured. To the best of
our knowledge, the results reported here are the rst of their
kind. Performing several experiments on the same MEMS
device repeatability of the test was assessed. In all loading
cases, the scatter of the data was small. By performing the
experiment on several switches, in the same wafer, the unifor-
mity of the manufacturing process was identied. This fea-
ture and the fact that the test is performed at the wafer level,
without the need of specially designed specimens and with a
commercially available nanoindenter, are expected to be very
valuable in quality control of MEMS devices in large-scale
manufacturing processes.
A major remaining challenge in the design of MEMS de-
vices is the capability to assess their reliability. For instance,
MEMS rf switches are expected to be actuated by a large
number of cycles in wireless or other applications. Problems
associated to stiction, relaxation of residual stresses that con-
trol the time response of the device, through a spring back ef-
fect, or even material fatigue can be envisioned. The method-
ology presented here can be employed to address some of
these problems. For instance, by electrostatically actuating
the large number of cycles of the switches and by performing
MDE experiments at periodic intervals, a complete evolution
of residual stress state and mechanical properties can be in-
vestigated. Complementary TEM, FIBand scanning electron
microscopy (SEM) studies performed on the tested devices
could provide valuable insight into failure mechanisms at
the microscale and, therefore, increase our ability to design
microdevices.
Acknowledgments
The authors acknowledge the support of Raytheon Sys-
tems Co., through a grant to Northwestern University. HDE
also acknowledges the support from NSF through a Young
Investigator Award, NSF-Career Award No CMS-9624364,
and FAA Award No DTFA03-01-C-00031. This work would
have not been possible without the input and rf switches pro-
vided by T. Baughn, S. Chen, and C. Goldsmith of Raytheon
Systems Co., Dallas, TX 75243. Special thanks are due to
Warren Oliver and Erik Herbert, MTS Systems Nano Instru-
ments Innovation Center, for providing valuable insight on
the experimental process using the Nanoindenter XP.
References
1. Geppert, L., RF Bridges to the Network, IEEE Spectrum Magazine,
6971, (January 2001).
2. Cass, S., Large Jobs for Little Devices, IEEE Spectrum Magazine,
7273, (January 2001).
3. Chalmers, P., Relay Races,Mechanical Engineering Magazine, 54
62 (January 2001).
4. Goldsmith, C., Lin, T., Powers, B., Wen, R., and Norvell, B., Mi-
cromechanical Membrane Switches for Microwave Applications, IEEE Mi-
crowave Theory Tech. Symp., 1, 9194 (1995).
5. Goldsmith, C., Yao, Z., Eshelman, S., and Denniston, D., Perfor-
mance of Low-loss RF MEMS Capacitive Switches, IEEE Microwave
Guided Wave Lett., 8 (8), 269271 (1998).
2003 Society for Experimental Mechanics Experimental Mechanics 315
6. Goldsmith, C., Malczewski, A., Yao, Z., Chen, S., Ehmke, J., andHinzel,
D., RF MEMS Variable Capacitors for Tunable Filters, Int. J. RF Mi-
crowave Computer-Aided Eng., 9 (4), 362374 (1999).
7. Goldsmith, C., Randall, J., Eshelman, S., Lin, T., Denniston, D., Chen,
S., and Norvell, B., Characteristics of Micromachined Switches at Mi-
crowave Frequencies, IEEE Microwave Theory Tech. Symp., 2, 11411144
(1996).
8. Small, M. and Nix, W., Analysis of the Accuracy of the Bulge Test in
Determining the Mechanical Properties of Thin Films, J. Mater. Res., 7 (6)
(1992).
9. Vlassak, J. and Nix, W., A New Bulge Test Technique for the Deter-
mination of Youngs Modulus and Poissons Ratio of Thin Films, J. Mater.
Res., 7 (12) (1992).
10. Small, M., Daniels, B., Clemens, B., and Nix, W., The Elastic Bi-
axial Modulus of AgPd Multilayered Thin Films Using the Bulge Test, J.
Mater. Res., 9 (1) (1994).
11. Yuan, B. and Sharpe Jr., W., Mechanical Testing of Polysilicon Thin
Films With the ISDG, EXPERIMENTAL TECHNIQUES, 21 (2), 3235
(1997).
12. Sharpe Jr., W., A New Technique for Measuring Poissons Ratio of
MEMS Materials, Proc. Mater. Res. Soc. Symp., 444, 185190, Boston, MA
(1996).
13. Sharpe Jr., W., ASME NADAI LectureElastoplastic Stress and
Strain Concentrations, J. Eng. Mater. Technol., 117, 17 (1995).
14. Zeng, H. and Sharpe Jr., W., A System for Measuring Biaxial Creep
Strains Over Short Gage Lengths, EXPERIMENTAL MECHANICS, 53,
8497 (1996).
15. Shell de Guzman, M., Neubauer, G., Flinn, P., and Nix, W., The Role
of Indentation on the Measured Hardness of Materials, Proc. Mater. Res.
Soc. Symp., 308, 613618 (1993).
16. Nix, W., Elastic and Plastic Properties of Thin Films on Substrates:
Nanoindentation Techniques, Mater. Sci. Eng. A, 234236, 3744 (1997).
17. Baker, S. and Nix, W., Mechanical Properties of Compositionally
Modulated AuNi Thin Films: Nanoindentation and Microcantilever Exper-
iments, J. Mater. Res., 9 (12), 31313144 (1994).
18. Tsui, T., Oliver, W., and Pharr, G., Inuences of Stress on the Mea-
surement of Mechanical Properties Using Nanoindentation: Part I. Exper-
imental Studies in an Aluminum Alloy, J. Mater. Res., 11 (3), 752759
(1996).
19. Tsui, T., Oliver, W., and Pharr, G., Inuences of Stress on the Mea-
surement of Mechanical Properties Using Nanoindentation: Part II. Finite
Element Simulations, J. Mater. Res., 11 (3), 760768 (1996).
20. Hong, S., Weihs, T.P., Bravman, J.C., and Nix, W.D., Measuring
Stiffness and Residual Stresses of Silicon Nitride Thin Film, J. Electron.
Mater., 19, 903909 (1990).
21. Fischer, M., MEMS Material Testing, Masters of Science, Purdue
University (1999).
22. Espinosa, H.D., Fischer, M., Herbert, E., and Oliver, W., Identi-
cation of Residual Stress State in an RF-MEMS Device, white paper, MTS
Systems Corporation (2001).
23. Espinosa, H.D., Fischer, M., Zhu, Y., and Lee, S. 3-D Computa-
tional Modeling of RF MEMS Switches, in Tech. Proc. 4th International
Conference on Modeling and Simulation of Microsystems, M. Laudon and
B. Romanowicz, eds., 402405 (2001).
24. Nix, W.D., Thin Film Reviews Mechanical Properties of Thin
Films, Metall. Trans. A, 20, 22172245 (1989).
25. Espinosa, H.D., Prorok, B.C., and Fischer, M., Elasticity, Plasticity
and Fracture of Thin Films and MEMS MaterialsPart I: A Novel Chip
Level Testing Methodology, J. Mech. Phys. Solids, submitted (2001).
26. Chen, S., Baughn, T.V., Yao, Z., and Goldsmith, C., A New In-situ
Residual Stress Measurement Method for MEMS Suspended Membranes,
J. Microelectromech. Syst., submitted (2001).
316 Vol. 43, No. 3, September 2003 2003 Society for Experimental Mechanics